Predictive Maintenance: Beyond The Crystal Ball
Industrial Maintenance & Plant Operation
Every manufacturing leader would love to know about a looming problem minutes, hours and even days before it occurs to prevent or minimize the negative effects of the issue. Yet, too often operations leaders still simply adopt the scheduled maintenance intervals recommended by equipment builders.
Generally, this is a safe policy, but generalized recommendations based on assumptions baked into equipment builders’ maintenance schedules can be wasteful or insufficient. One manufacturer may push equipment speed, load, or duty cycle to meet its delivery commitments while another part maker may use a machine less strenuously.
Predictive maintenance (PdM) has emerged as a viable option for manufacturing companies of all sizes to not only avoid unplanned downtime but also extend the life of equipment. By using a PdM system to detect, predict and avoid one production disruption per calendar quarter, a manufacturer can save tens to hundreds of thousands of dollars in overtime, expedited freight for product shipments, and rush order fees for repair parts.
Predictive maintenance is powered by smart manufacturing technology, also known as the Industrial Internet of Things (IIoT), which incorporates real-time measurements of equipment and tooling performance via sensors and related machine controllers. The data then is fed in real time into analysis and reporting software. Through this technology, manufacturers can stay connected to the shop floor and gain deeper insights into their operations.
For years, there have been IIoT sensors for machine speed, position, temperature, pressure, electrical current and voltage, and flow with imbedded low-power electronics to provide intelligence and various forms of communications abilities. However, only recently have prices for sensors dropped to the point that it is increasingly practical for large and small manufacturers, alike, to instrument older machines and make them “smarter” to provide real-time data. This helps to address the mix of well-used iron horses running on the floor alongside newer, highly automated equipment run by sophisticated controllers that enable a complete “smart” view of the entire operation.
At the same time, the software used to capture, roll up, store and analyze the stream of operations data, often called a manufacturing execution system (MES), has matured. Together, IIoT and MES technology enable process engineers and maintenance staff to work with information technology (IT) and/or third-party providers in building a capable predictive maintenance system.
Before implementing a predictive management system to improve equipment availability and efficiency, though, plant managers and their maintenance teams need to ensure that they address three key real-time functions: monitoring, analysis and feedback.
Sensing a drop in pneumatic or hydraulic pressure, fluctuation in motor current, variations in shaft rotational speed as an unusual vibration, or a rise in temperature can be valuable tools in predicting and preventing production disruptions. Monitoring of such events can be accomplished by using simple sensors, smart sensors, or the equipment’s own controller to gather data and analyze it immediately.
When individual work centers have machine controls with built-in sensors, they need to be connected via a shop floor network to create a comprehensive view of operations. This network may consist of hardwired cables or wireless communications, which transmit measurements to a central “edge” server in the plant that aggregates the data. If a plant does not already have a network in place, it is now relatively simple to create one. In either case, when connecting to existing controllers, it is important to identify the communications protocol required for access to internal “tags” (information addresses or “registers” in the controller) that temporarily store the latest measurement data.
Many older machines either do not have an integrated electronic controller, or the controller does not capture relevant measurements. However, many manufacturers simply add sensors to their tooling and process machines. Then, they connect the additional sensors to dedicated data gathering/concentrating devices, which in turn communicate the data to a local or remote server for analysis and reporting.
Once communications are established with the plant equipment, one of two actions will occur: Relevant measurements may be stored locally for immediate trend analysis, so metrics can be reported line-side via tablets and monitors. Alternatively, measurements may be transferred to an enterprise database on-premises or in the cloud for further processing. Either way, continuously-updated trend analysis and metrics provide key staff with leading indicators of looming mechanical or electrical failures that would be disruptive and costly to repair.
For many years, analysis has been a manual process performed by experienced production workers and skilled maintenance techs. These employees can identify when vibrations, hesitations, odors, buzzes, rattles and squeaks are signaling that production equipment is on the verge of failure. As these employees retire, this skill is leaving the factory with them.
Increasingly, manufacturers are turning to software for analysis of performance measurements. Usually, this software looks for patterns of irregularities or statistical trends that indicate a future loss of control before a major quality problem or catastrophic equipment failure that disrupts production occurs. During the initial setup, most analysis systems depend on both the experience of plant staff and machine operating manuals to set up the high and low limits of acceptable performance.
An effective technology for analyzing and reporting unhealthy trends in measurements is statistical process control (SPC) logic. This business logic is available with standalone SPC software, and it is often incorporated into modern MES and enterprise resource planning (ERP) software. Real-time reporting through SPC software connected with modules for quality, scheduling and maintenance helps to ensure that plant employees have access to actionable information.
Timely access to accurate metrics enables manufacturers to complete touch-ups, adjustments, or minor repairs between shifts, so uptime is maximized. Additionally, insights from real-time analytics facilitate the scheduling of cleaning, lubing or replacement of worn components closer to their point of likely failure based on accurate fact-based data, extending the useful life of valuable production process equipment.
Machine components, such as specialty seals, precision bearings and sensitive valves, can be hard to reach and very expensive. Often, teams find that these maintenance repair and overhaul (MRO) items are still in acceptable operating condition during preventative scheduled teardowns — leaving the company with a loss of precious production capacity, wasted time for skilled techs and cash unnecessarily tied up in inventory.
By contrast, feedback via real-time, predictive analysis gives employees a factual basis for managing asset “health” and ultimately production risk — empowering them to set application-appropriate maintenance schedules for equipment, depending on how they are used in a given plant. By relying on automated process digital dashboards and alerts, instead of exclusively using planned preventative maintenance intervals, manufacturers can maximize uptime and minimize the effort and costs needed to keep the factory humming along. In doing so, maintenance managers and technicians can apply the risk-based decision-making that is encouraged by the ISO 9001:2015 quality standard.
Of course, feedback is only useful if it is recognized, and on hectic factory floors, employees may not notice a change in operating conditions, even if it is reported on displays throughout the facility. This issue can be addressed by enabling automated alerts in the PdM system to capture employees’ attention and assure that negative conditions are not missed.
For example, when a pending issue is sensed, MES software can be set to trigger an email or text warning to key staff members. Alerts also can be signaled via a light pole next to each machine, or they may take the form of automated announcements on the plant’s public-address system.
As operations leaders weigh the significant potential benefits for maintenance teams, it might seem intimidating to put a system of predictive maintenance in place. However, innovators who have successfully implemented PdM technology provide best practices that other manufacturers can apply:
- Start with the end in mind, and identify specific priority work centers.
- Pilot one or two work centers in each generation to gain experience and team member buy-in.
- Keep the project scope simple to start.
- Analyze the return on investment (ROI) from each pilot project based on the actual costs and benefits versus planned.
- Capture lessons learned to apply in the future.
- Create an investment and roll-out plan to execute the predictive maintenance effort in phases.
By thoughtfully planning and implementing equipment measuring hardware and monitoring software systems, manufacturers are at the brink of a new, era of intelligently protecting production capacity, minimizing costs and managing precious assets efficiently.
The information contained in this publication is intended for general information purposes. No representation is made that the information or references are complete or remain current. Click here for Grainger's full legal disclaimer.